Analysis of appropriate spectral similarity methods to classify target species using CASI hyperspectral images

This study analyzed various spectral methods to classify target spectra from airborne hyperspectral images. Various spectral similarity methods have been developed to match target signatures with the reference spectra for various applications. Spectral similarity measures are effective in classifying spectral signatures because they can mitigate illumination-change effects. However, there has been little emphasis on the use of appropriate spectral similarity measures to classify specific target signatures, such as of the corn, potato, and lotus, from airborne hyperspectral images. In this study, field data obtained from field spectrocopy and a Compact Airborne Spectrographic Imager (CASI) containing various target spectra were used to select appropriate spectral similarity methods. Not only original measures such as Euclidean distance (ED), spectral angle mapper (SAM), spectral correlation angle (SCA), and spectral information divergence (SID), but also newly developed hybrid methods, were evaluated to test the ability to classify specific target spectra. The performance of spectral measures was evaluated by the probability of spectral discrimination (PSD) and the power of spectral discrimination (PWSD).